1 | """ |
---|
2 | Sasview model constructor. |
---|
3 | |
---|
4 | Given a module defining an OpenCL kernel such as sasmodels.models.cylinder, |
---|
5 | create a sasview model class to run that kernel as follows:: |
---|
6 | |
---|
7 | from sasmodels.sasview_model import load_custom_model |
---|
8 | CylinderModel = load_custom_model('sasmodels/models/cylinder.py') |
---|
9 | """ |
---|
10 | from __future__ import print_function |
---|
11 | |
---|
12 | import math |
---|
13 | from copy import deepcopy |
---|
14 | import collections |
---|
15 | import traceback |
---|
16 | import logging |
---|
17 | from os.path import basename, splitext, abspath, getmtime |
---|
18 | try: |
---|
19 | import _thread as thread |
---|
20 | except ImportError: |
---|
21 | import thread |
---|
22 | |
---|
23 | import numpy as np # type: ignore |
---|
24 | |
---|
25 | from . import core |
---|
26 | from . import custom |
---|
27 | from . import kernelcl |
---|
28 | from . import product |
---|
29 | from . import generate |
---|
30 | from . import weights |
---|
31 | from . import modelinfo |
---|
32 | from .details import make_kernel_args, dispersion_mesh |
---|
33 | from .kernelcl import reset_environment |
---|
34 | |
---|
35 | # pylint: disable=unused-import |
---|
36 | try: |
---|
37 | from typing import (Dict, Mapping, Any, Sequence, Tuple, NamedTuple, |
---|
38 | List, Optional, Union, Callable) |
---|
39 | from .modelinfo import ModelInfo, Parameter |
---|
40 | from .kernel import KernelModel |
---|
41 | MultiplicityInfoType = NamedTuple( |
---|
42 | 'MultiplicityInfo', |
---|
43 | [("number", int), ("control", str), ("choices", List[str]), |
---|
44 | ("x_axis_label", str)]) |
---|
45 | SasviewModelType = Callable[[int], "SasviewModel"] |
---|
46 | except ImportError: |
---|
47 | pass |
---|
48 | # pylint: enable=unused-import |
---|
49 | |
---|
50 | logger = logging.getLogger(__name__) |
---|
51 | |
---|
52 | calculation_lock = thread.allocate_lock() |
---|
53 | |
---|
54 | #: True if pre-existing plugins, with the old names and parameters, should |
---|
55 | #: continue to be supported. |
---|
56 | SUPPORT_OLD_STYLE_PLUGINS = True |
---|
57 | |
---|
58 | # TODO: separate x_axis_label from multiplicity info |
---|
59 | MultiplicityInfo = collections.namedtuple( |
---|
60 | 'MultiplicityInfo', |
---|
61 | ["number", "control", "choices", "x_axis_label"], |
---|
62 | ) |
---|
63 | |
---|
64 | #: set of defined models (standard and custom) |
---|
65 | MODELS = {} # type: Dict[str, SasviewModelType] |
---|
66 | # TODO: remove unused MODEL_BY_PATH cache once sasview no longer references it |
---|
67 | #: custom model {path: model} mapping so we can check timestamps |
---|
68 | MODEL_BY_PATH = {} # type: Dict[str, SasviewModelType] |
---|
69 | #: Track modules that we have loaded so we can determine whether the model |
---|
70 | #: has changed since we last reloaded. |
---|
71 | _CACHED_MODULE = {} # type: Dict[str, "module"] |
---|
72 | |
---|
73 | def reset_environment(): |
---|
74 | # type: () -> None |
---|
75 | """ |
---|
76 | Clear the compute engine context so that the GUI can change devices. |
---|
77 | |
---|
78 | This removes all compiled kernels, even those that are active on fit |
---|
79 | pages, but they will be restored the next time they are needed. |
---|
80 | """ |
---|
81 | kernelcl.reset_environment() |
---|
82 | for model in MODELS.values(): |
---|
83 | model._model = None |
---|
84 | |
---|
85 | def find_model(modelname): |
---|
86 | # type: (str) -> SasviewModelType |
---|
87 | """ |
---|
88 | Find a model by name. If the model name ends in py, try loading it from |
---|
89 | custom models, otherwise look for it in the list of builtin models. |
---|
90 | """ |
---|
91 | # TODO: used by sum/product model to load an existing model |
---|
92 | # TODO: doesn't handle custom models properly |
---|
93 | if modelname.endswith('.py'): |
---|
94 | return load_custom_model(modelname) |
---|
95 | elif modelname in MODELS: |
---|
96 | return MODELS[modelname] |
---|
97 | else: |
---|
98 | raise ValueError("unknown model %r"%modelname) |
---|
99 | |
---|
100 | |
---|
101 | # TODO: figure out how to say that the return type is a subclass |
---|
102 | def load_standard_models(): |
---|
103 | # type: () -> List[SasviewModelType] |
---|
104 | """ |
---|
105 | Load and return the list of predefined models. |
---|
106 | |
---|
107 | If there is an error loading a model, then a traceback is logged and the |
---|
108 | model is not returned. |
---|
109 | """ |
---|
110 | for name in core.list_models(): |
---|
111 | try: |
---|
112 | MODELS[name] = _make_standard_model(name) |
---|
113 | except Exception: |
---|
114 | logger.error(traceback.format_exc()) |
---|
115 | if SUPPORT_OLD_STYLE_PLUGINS: |
---|
116 | _register_old_models() |
---|
117 | |
---|
118 | return list(MODELS.values()) |
---|
119 | |
---|
120 | |
---|
121 | def load_custom_model(path): |
---|
122 | # type: (str) -> SasviewModelType |
---|
123 | """ |
---|
124 | Load a custom model given the model path. |
---|
125 | """ |
---|
126 | #logger.info("Loading model %s", path) |
---|
127 | |
---|
128 | # Load the kernel module. This may already be cached by the loader, so |
---|
129 | # only requires checking the timestamps of the dependents. |
---|
130 | kernel_module = custom.load_custom_kernel_module(path) |
---|
131 | |
---|
132 | # Check if the module has changed since we last looked. |
---|
133 | reloaded = kernel_module != _CACHED_MODULE.get(path, None) |
---|
134 | _CACHED_MODULE[path] = kernel_module |
---|
135 | |
---|
136 | # Turn the module into a model. We need to do this in even if the |
---|
137 | # model has already been loaded so that we can determine the model |
---|
138 | # name and retrieve it from the MODELS cache. |
---|
139 | model = getattr(kernel_module, 'Model', None) |
---|
140 | if model is not None: |
---|
141 | # Old style models do not set the name in the class attributes, so |
---|
142 | # set it here; this name will be overridden when the object is created |
---|
143 | # with an instance variable that has the same value. |
---|
144 | if model.name == "": |
---|
145 | model.name = splitext(basename(path))[0] |
---|
146 | if not hasattr(model, 'filename'): |
---|
147 | model.filename = abspath(kernel_module.__file__).replace('.pyc', '.py') |
---|
148 | if not hasattr(model, 'id'): |
---|
149 | model.id = splitext(basename(model.filename))[0] |
---|
150 | else: |
---|
151 | model_info = modelinfo.make_model_info(kernel_module) |
---|
152 | model = make_model_from_info(model_info) |
---|
153 | |
---|
154 | # If a model name already exists and we are loading a different model, |
---|
155 | # use the model file name as the model name. |
---|
156 | if model.name in MODELS and not model.filename == MODELS[model.name].filename: |
---|
157 | _previous_name = model.name |
---|
158 | model.name = model.id |
---|
159 | |
---|
160 | # If the new model name is still in the model list (for instance, |
---|
161 | # if we put a cylinder.py in our plug-in directory), then append |
---|
162 | # an identifier. |
---|
163 | if model.name in MODELS and not model.filename == MODELS[model.name].filename: |
---|
164 | model.name = model.id + '_user' |
---|
165 | logger.info("Model %s already exists: using %s [%s]", |
---|
166 | _previous_name, model.name, model.filename) |
---|
167 | |
---|
168 | # Only update the model if the module has changed |
---|
169 | if reloaded or model.name not in MODELS: |
---|
170 | MODELS[model.name] = model |
---|
171 | |
---|
172 | return MODELS[model.name] |
---|
173 | |
---|
174 | |
---|
175 | def make_model_from_info(model_info): |
---|
176 | # type: (ModelInfo) -> SasviewModelType |
---|
177 | """ |
---|
178 | Convert *model_info* into a SasView model wrapper. |
---|
179 | """ |
---|
180 | def __init__(self, multiplicity=None): |
---|
181 | SasviewModel.__init__(self, multiplicity=multiplicity) |
---|
182 | attrs = _generate_model_attributes(model_info) |
---|
183 | attrs['__init__'] = __init__ |
---|
184 | attrs['filename'] = model_info.filename |
---|
185 | ConstructedModel = type(model_info.name, (SasviewModel,), attrs) # type: SasviewModelType |
---|
186 | return ConstructedModel |
---|
187 | |
---|
188 | |
---|
189 | def _make_standard_model(name): |
---|
190 | # type: (str) -> SasviewModelType |
---|
191 | """ |
---|
192 | Load the sasview model defined by *name*. |
---|
193 | |
---|
194 | *name* can be a standard model name or a path to a custom model. |
---|
195 | |
---|
196 | Returns a class that can be used directly as a sasview model. |
---|
197 | """ |
---|
198 | kernel_module = generate.load_kernel_module(name) |
---|
199 | model_info = modelinfo.make_model_info(kernel_module) |
---|
200 | return make_model_from_info(model_info) |
---|
201 | |
---|
202 | |
---|
203 | def _register_old_models(): |
---|
204 | # type: () -> None |
---|
205 | """ |
---|
206 | Place the new models into sasview under the old names. |
---|
207 | |
---|
208 | Monkey patch sas.sascalc.fit as sas.models so that sas.models.pluginmodel |
---|
209 | is available to the plugin modules. |
---|
210 | """ |
---|
211 | import sys |
---|
212 | import sas # needed in order to set sas.models |
---|
213 | import sas.sascalc.fit |
---|
214 | sys.modules['sas.models'] = sas.sascalc.fit |
---|
215 | sas.models = sas.sascalc.fit |
---|
216 | import sas.models |
---|
217 | from sasmodels.conversion_table import CONVERSION_TABLE |
---|
218 | |
---|
219 | for new_name, conversion in CONVERSION_TABLE.get((3, 1, 2), {}).items(): |
---|
220 | # CoreShellEllipsoidModel => core_shell_ellipsoid:1 |
---|
221 | new_name = new_name.split(':')[0] |
---|
222 | old_name = conversion[0] if len(conversion) < 3 else conversion[2] |
---|
223 | module_attrs = {old_name: find_model(new_name)} |
---|
224 | ConstructedModule = type(old_name, (), module_attrs) |
---|
225 | old_path = 'sas.models.' + old_name |
---|
226 | setattr(sas.models, old_path, ConstructedModule) |
---|
227 | sys.modules[old_path] = ConstructedModule |
---|
228 | |
---|
229 | |
---|
230 | def MultiplicationModel(form_factor, structure_factor): |
---|
231 | # type: ("SasviewModel", "SasviewModel") -> "SasviewModel" |
---|
232 | """ |
---|
233 | Returns a constructed product model from form_factor and structure_factor. |
---|
234 | """ |
---|
235 | model_info = product.make_product_info(form_factor._model_info, |
---|
236 | structure_factor._model_info) |
---|
237 | ConstructedModel = make_model_from_info(model_info) |
---|
238 | return ConstructedModel(form_factor.multiplicity) |
---|
239 | |
---|
240 | |
---|
241 | def _generate_model_attributes(model_info): |
---|
242 | # type: (ModelInfo) -> Dict[str, Any] |
---|
243 | """ |
---|
244 | Generate the class attributes for the model. |
---|
245 | |
---|
246 | This should include all the information necessary to query the model |
---|
247 | details so that you do not need to instantiate a model to query it. |
---|
248 | |
---|
249 | All the attributes should be immutable to avoid accidents. |
---|
250 | """ |
---|
251 | |
---|
252 | # TODO: allow model to override axis labels input/output name/unit |
---|
253 | |
---|
254 | # Process multiplicity |
---|
255 | non_fittable = [] # type: List[str] |
---|
256 | xlabel = model_info.profile_axes[0] if model_info.profile is not None else "" |
---|
257 | variants = MultiplicityInfo(0, "", [], xlabel) |
---|
258 | for p in model_info.parameters.kernel_parameters: |
---|
259 | if p.name == model_info.control: |
---|
260 | non_fittable.append(p.name) |
---|
261 | variants = MultiplicityInfo( |
---|
262 | len(p.choices) if p.choices else int(p.limits[1]), |
---|
263 | p.name, p.choices, xlabel |
---|
264 | ) |
---|
265 | break |
---|
266 | |
---|
267 | # Only a single drop-down list parameter available |
---|
268 | fun_list = [] |
---|
269 | for p in model_info.parameters.kernel_parameters: |
---|
270 | if p.choices: |
---|
271 | fun_list = p.choices |
---|
272 | if p.length > 1: |
---|
273 | non_fittable.extend(p.id+str(k) for k in range(1, p.length+1)) |
---|
274 | break |
---|
275 | |
---|
276 | # Organize parameter sets |
---|
277 | orientation_params = [] |
---|
278 | magnetic_params = [] |
---|
279 | fixed = [] |
---|
280 | for p in model_info.parameters.user_parameters({}, is2d=True): |
---|
281 | if p.type == 'orientation': |
---|
282 | orientation_params.append(p.name) |
---|
283 | orientation_params.append(p.name+".width") |
---|
284 | fixed.append(p.name+".width") |
---|
285 | elif p.type == 'magnetic': |
---|
286 | orientation_params.append(p.name) |
---|
287 | magnetic_params.append(p.name) |
---|
288 | fixed.append(p.name+".width") |
---|
289 | |
---|
290 | |
---|
291 | # Build class dictionary |
---|
292 | attrs = {} # type: Dict[str, Any] |
---|
293 | attrs['_model_info'] = model_info |
---|
294 | attrs['name'] = model_info.name |
---|
295 | attrs['id'] = model_info.id |
---|
296 | attrs['description'] = model_info.description |
---|
297 | attrs['category'] = model_info.category |
---|
298 | attrs['is_structure_factor'] = model_info.structure_factor |
---|
299 | attrs['is_form_factor'] = model_info.effective_radius_type is not None |
---|
300 | attrs['is_multiplicity_model'] = variants[0] > 1 |
---|
301 | attrs['multiplicity_info'] = variants |
---|
302 | attrs['orientation_params'] = tuple(orientation_params) |
---|
303 | attrs['magnetic_params'] = tuple(magnetic_params) |
---|
304 | attrs['fixed'] = tuple(fixed) |
---|
305 | attrs['non_fittable'] = tuple(non_fittable) |
---|
306 | attrs['fun_list'] = tuple(fun_list) |
---|
307 | |
---|
308 | return attrs |
---|
309 | |
---|
310 | class SasviewModel(object): |
---|
311 | """ |
---|
312 | Sasview wrapper for opencl/ctypes model. |
---|
313 | """ |
---|
314 | # Model parameters for the specific model are set in the class constructor |
---|
315 | # via the _generate_model_attributes function, which subclasses |
---|
316 | # SasviewModel. They are included here for typing and documentation |
---|
317 | # purposes. |
---|
318 | _model = None # type: KernelModel |
---|
319 | _model_info = None # type: ModelInfo |
---|
320 | #: load/save name for the model |
---|
321 | id = None # type: str |
---|
322 | #: display name for the model |
---|
323 | name = None # type: str |
---|
324 | #: short model description |
---|
325 | description = None # type: str |
---|
326 | #: default model category |
---|
327 | category = None # type: str |
---|
328 | |
---|
329 | #: names of the orientation parameters in the order they appear |
---|
330 | orientation_params = None # type: List[str] |
---|
331 | #: names of the magnetic parameters in the order they appear |
---|
332 | magnetic_params = None # type: List[str] |
---|
333 | #: names of the fittable parameters |
---|
334 | fixed = None # type: List[str] |
---|
335 | # TODO: the attribute fixed is ill-named |
---|
336 | |
---|
337 | # Axis labels |
---|
338 | input_name = "Q" |
---|
339 | input_unit = "A^{-1}" |
---|
340 | output_name = "Intensity" |
---|
341 | output_unit = "cm^{-1}" |
---|
342 | |
---|
343 | #: default cutoff for polydispersity |
---|
344 | cutoff = 1e-5 |
---|
345 | |
---|
346 | # Note: Use non-mutable values for class attributes to avoid errors |
---|
347 | #: parameters that are not fitted |
---|
348 | non_fittable = () # type: Sequence[str] |
---|
349 | |
---|
350 | #: True if model should appear as a structure factor |
---|
351 | is_structure_factor = False |
---|
352 | #: True if model should appear as a form factor |
---|
353 | is_form_factor = False |
---|
354 | #: True if model has multiplicity |
---|
355 | is_multiplicity_model = False |
---|
356 | #: Multiplicity information |
---|
357 | multiplicity_info = None # type: MultiplicityInfoType |
---|
358 | |
---|
359 | # Per-instance variables |
---|
360 | #: parameter {name: value} mapping |
---|
361 | params = None # type: Dict[str, float] |
---|
362 | #: values for dispersion width, npts, nsigmas and type |
---|
363 | dispersion = None # type: Dict[str, Any] |
---|
364 | #: units and limits for each parameter |
---|
365 | details = None # type: Dict[str, Sequence[Any]] |
---|
366 | # # actual type is Dict[str, List[str, float, float]] |
---|
367 | #: multiplicity value, or None if no multiplicity on the model |
---|
368 | multiplicity = None # type: Optional[int] |
---|
369 | #: memory for polydispersity array if using ArrayDispersion (used by sasview). |
---|
370 | _persistency_dict = None # type: Dict[str, Tuple[np.ndarray, np.ndarray]] |
---|
371 | |
---|
372 | def __init__(self, multiplicity=None): |
---|
373 | # type: (Optional[int]) -> None |
---|
374 | |
---|
375 | # TODO: _persistency_dict to persistency_dict throughout sasview |
---|
376 | # TODO: refactor multiplicity to encompass variants |
---|
377 | # TODO: dispersion should be a class |
---|
378 | # TODO: refactor multiplicity info |
---|
379 | # TODO: separate profile view from multiplicity |
---|
380 | # The button label, x and y axis labels and scale need to be under |
---|
381 | # the control of the model, not the fit page. Maximum flexibility, |
---|
382 | # the fit page would supply the canvas and the profile could plot |
---|
383 | # how it wants, but this assumes matplotlib. Next level is that |
---|
384 | # we provide some sort of data description including title, labels |
---|
385 | # and lines to plot. |
---|
386 | |
---|
387 | # Get the list of hidden parameters given the multiplicity |
---|
388 | # Don't include multiplicity in the list of parameters |
---|
389 | self.multiplicity = multiplicity |
---|
390 | if multiplicity is not None: |
---|
391 | hidden = self._model_info.get_hidden_parameters(multiplicity) |
---|
392 | hidden |= set([self.multiplicity_info.control]) |
---|
393 | else: |
---|
394 | hidden = set() |
---|
395 | if self._model_info.structure_factor: |
---|
396 | hidden.add('scale') |
---|
397 | hidden.add('background') |
---|
398 | |
---|
399 | # Update the parameter lists to exclude any hidden parameters |
---|
400 | self.magnetic_params = tuple(pname for pname in self.magnetic_params |
---|
401 | if pname not in hidden) |
---|
402 | self.orientation_params = tuple(pname for pname in self.orientation_params |
---|
403 | if pname not in hidden) |
---|
404 | |
---|
405 | self._persistency_dict = {} |
---|
406 | self.params = collections.OrderedDict() |
---|
407 | self.dispersion = collections.OrderedDict() |
---|
408 | self.details = {} |
---|
409 | for p in self._model_info.parameters.user_parameters({}, is2d=True): |
---|
410 | if p.name in hidden: |
---|
411 | continue |
---|
412 | self.params[p.name] = p.default |
---|
413 | self.details[p.id] = [p.units, p.limits[0], p.limits[1]] |
---|
414 | if p.polydisperse: |
---|
415 | self.details[p.id+".width"] = [ |
---|
416 | "", 0.0, 1.0 if p.relative_pd else np.inf |
---|
417 | ] |
---|
418 | self.dispersion[p.name] = { |
---|
419 | 'width': 0, |
---|
420 | 'npts': 35, |
---|
421 | 'nsigmas': 3, |
---|
422 | 'type': 'gaussian', |
---|
423 | } |
---|
424 | |
---|
425 | def __get_state__(self): |
---|
426 | # type: () -> Dict[str, Any] |
---|
427 | state = self.__dict__.copy() |
---|
428 | state.pop('_model') |
---|
429 | # May need to reload model info on set state since it has pointers |
---|
430 | # to python implementations of Iq, etc. |
---|
431 | #state.pop('_model_info') |
---|
432 | return state |
---|
433 | |
---|
434 | def __set_state__(self, state): |
---|
435 | # type: (Dict[str, Any]) -> None |
---|
436 | self.__dict__ = state |
---|
437 | self._model = None |
---|
438 | |
---|
439 | def __str__(self): |
---|
440 | # type: () -> str |
---|
441 | """ |
---|
442 | :return: string representation |
---|
443 | """ |
---|
444 | return self.name |
---|
445 | |
---|
446 | def is_fittable(self, par_name): |
---|
447 | # type: (str) -> bool |
---|
448 | """ |
---|
449 | Check if a given parameter is fittable or not |
---|
450 | |
---|
451 | :param par_name: the parameter name to check |
---|
452 | """ |
---|
453 | return par_name in self.fixed |
---|
454 | #For the future |
---|
455 | #return self.params[str(par_name)].is_fittable() |
---|
456 | |
---|
457 | |
---|
458 | def getProfile(self): |
---|
459 | # type: () -> (np.ndarray, np.ndarray) |
---|
460 | """ |
---|
461 | Get SLD profile |
---|
462 | |
---|
463 | : return: (z, beta) where z is a list of depth of the transition points |
---|
464 | beta is a list of the corresponding SLD values |
---|
465 | """ |
---|
466 | args = {} # type: Dict[str, Any] |
---|
467 | for p in self._model_info.parameters.kernel_parameters: |
---|
468 | if p.id == self.multiplicity_info.control: |
---|
469 | value = float(self.multiplicity) |
---|
470 | elif p.length == 1: |
---|
471 | value = self.params.get(p.id, np.NaN) |
---|
472 | else: |
---|
473 | value = np.array([self.params.get(p.id+str(k), np.NaN) |
---|
474 | for k in range(1, p.length+1)]) |
---|
475 | args[p.id] = value |
---|
476 | |
---|
477 | x, y = self._model_info.profile(**args) |
---|
478 | return x, 1e-6*y |
---|
479 | |
---|
480 | def setParam(self, name, value): |
---|
481 | # type: (str, float) -> None |
---|
482 | """ |
---|
483 | Set the value of a model parameter |
---|
484 | |
---|
485 | :param name: name of the parameter |
---|
486 | :param value: value of the parameter |
---|
487 | |
---|
488 | """ |
---|
489 | # Look for dispersion parameters |
---|
490 | toks = name.split('.') |
---|
491 | if len(toks) == 2: |
---|
492 | for item in self.dispersion.keys(): |
---|
493 | if item == toks[0]: |
---|
494 | for par in self.dispersion[item]: |
---|
495 | if par == toks[1]: |
---|
496 | self.dispersion[item][par] = value |
---|
497 | return |
---|
498 | else: |
---|
499 | # Look for standard parameter |
---|
500 | for item in self.params.keys(): |
---|
501 | if item == name: |
---|
502 | self.params[item] = value |
---|
503 | return |
---|
504 | |
---|
505 | raise ValueError("Model does not contain parameter %s" % name) |
---|
506 | |
---|
507 | def getParam(self, name): |
---|
508 | # type: (str) -> float |
---|
509 | """ |
---|
510 | Set the value of a model parameter |
---|
511 | |
---|
512 | :param name: name of the parameter |
---|
513 | |
---|
514 | """ |
---|
515 | # Look for dispersion parameters |
---|
516 | toks = name.split('.') |
---|
517 | if len(toks) == 2: |
---|
518 | for item in self.dispersion.keys(): |
---|
519 | if item == toks[0]: |
---|
520 | for par in self.dispersion[item]: |
---|
521 | if par == toks[1]: |
---|
522 | return self.dispersion[item][par] |
---|
523 | else: |
---|
524 | # Look for standard parameter |
---|
525 | for item in self.params.keys(): |
---|
526 | if item == name: |
---|
527 | return self.params[item] |
---|
528 | |
---|
529 | raise ValueError("Model does not contain parameter %s" % name) |
---|
530 | |
---|
531 | def getParamList(self): |
---|
532 | # type: () -> Sequence[str] |
---|
533 | """ |
---|
534 | Return a list of all available parameters for the model |
---|
535 | """ |
---|
536 | param_list = list(self.params.keys()) |
---|
537 | # WARNING: Extending the list with the dispersion parameters |
---|
538 | param_list.extend(self.getDispParamList()) |
---|
539 | return param_list |
---|
540 | |
---|
541 | def getDispParamList(self): |
---|
542 | # type: () -> Sequence[str] |
---|
543 | """ |
---|
544 | Return a list of polydispersity parameters for the model |
---|
545 | """ |
---|
546 | # TODO: fix test so that parameter order doesn't matter |
---|
547 | ret = ['%s.%s' % (p_name, ext) |
---|
548 | for p_name in self.dispersion.keys() |
---|
549 | for ext in ('npts', 'nsigmas', 'width')] |
---|
550 | #print(ret) |
---|
551 | return ret |
---|
552 | |
---|
553 | def clone(self): |
---|
554 | # type: () -> "SasviewModel" |
---|
555 | """ Return a identical copy of self """ |
---|
556 | return deepcopy(self) |
---|
557 | |
---|
558 | def run(self, x=0.0): |
---|
559 | # type: (Union[float, (float, float), List[float]]) -> float |
---|
560 | """ |
---|
561 | Evaluate the model |
---|
562 | |
---|
563 | :param x: input q, or [q,phi] |
---|
564 | |
---|
565 | :return: scattering function P(q) |
---|
566 | |
---|
567 | **DEPRECATED**: use calculate_Iq instead |
---|
568 | """ |
---|
569 | if isinstance(x, (list, tuple)): |
---|
570 | # pylint: disable=unpacking-non-sequence |
---|
571 | q, phi = x |
---|
572 | result, _ = self.calculate_Iq([q*math.cos(phi)], [q*math.sin(phi)]) |
---|
573 | return result[0] |
---|
574 | else: |
---|
575 | result, _ = self.calculate_Iq([x]) |
---|
576 | return result[0] |
---|
577 | |
---|
578 | |
---|
579 | def runXY(self, x=0.0): |
---|
580 | # type: (Union[float, (float, float), List[float]]) -> float |
---|
581 | """ |
---|
582 | Evaluate the model in cartesian coordinates |
---|
583 | |
---|
584 | :param x: input q, or [qx, qy] |
---|
585 | |
---|
586 | :return: scattering function P(q) |
---|
587 | |
---|
588 | **DEPRECATED**: use calculate_Iq instead |
---|
589 | """ |
---|
590 | if isinstance(x, (list, tuple)): |
---|
591 | result, _ = self.calculate_Iq([x[0]], [x[1]]) |
---|
592 | return result[0] |
---|
593 | else: |
---|
594 | result, _ = self.calculate_Iq([x]) |
---|
595 | return result[0] |
---|
596 | |
---|
597 | def evalDistribution(self, qdist): |
---|
598 | # type: (Union[np.ndarray, Tuple[np.ndarray, np.ndarray], List[np.ndarray]]) -> np.ndarray |
---|
599 | r""" |
---|
600 | Evaluate a distribution of q-values. |
---|
601 | |
---|
602 | :param qdist: array of q or a list of arrays [qx,qy] |
---|
603 | |
---|
604 | * For 1D, a numpy array is expected as input |
---|
605 | |
---|
606 | :: |
---|
607 | |
---|
608 | evalDistribution(q) |
---|
609 | |
---|
610 | where *q* is a numpy array. |
---|
611 | |
---|
612 | * For 2D, a list of *[qx,qy]* is expected with 1D arrays as input |
---|
613 | |
---|
614 | :: |
---|
615 | |
---|
616 | qx = [ qx[0], qx[1], qx[2], ....] |
---|
617 | qy = [ qy[0], qy[1], qy[2], ....] |
---|
618 | |
---|
619 | If the model is 1D only, then |
---|
620 | |
---|
621 | .. math:: |
---|
622 | |
---|
623 | q = \sqrt{q_x^2+q_y^2} |
---|
624 | |
---|
625 | """ |
---|
626 | if isinstance(qdist, (list, tuple)): |
---|
627 | # Check whether we have a list of ndarrays [qx,qy] |
---|
628 | qx, qy = qdist |
---|
629 | result, _ = self.calculate_Iq(qx, qy) |
---|
630 | return result |
---|
631 | |
---|
632 | elif isinstance(qdist, np.ndarray): |
---|
633 | # We have a simple 1D distribution of q-values |
---|
634 | result, _ = self.calculate_Iq(qdist) |
---|
635 | return result |
---|
636 | |
---|
637 | else: |
---|
638 | raise TypeError("evalDistribution expects q or [qx, qy], not %r" |
---|
639 | % type(qdist)) |
---|
640 | |
---|
641 | def calc_composition_models(self, qx): |
---|
642 | """ |
---|
643 | returns parts of the composition model or None if not a composition |
---|
644 | model. |
---|
645 | """ |
---|
646 | # TODO: have calculate_Iq return the intermediates. |
---|
647 | # |
---|
648 | # The current interface causes calculate_Iq() to be called twice, |
---|
649 | # once to get the combined result and again to get the intermediate |
---|
650 | # results. This is necessary for now. |
---|
651 | # Long term, the solution is to change the interface to calculate_Iq |
---|
652 | # so that it returns a results object containing all the bits: |
---|
653 | # the A, B, C, ... of the composition model (and any subcomponents?) |
---|
654 | # the P and S of the product model |
---|
655 | # the combined model before resolution smearing, |
---|
656 | # the sasmodel before sesans conversion, |
---|
657 | # the oriented 2D model used to fit oriented usans data, |
---|
658 | # the final I(q), |
---|
659 | # ... |
---|
660 | # |
---|
661 | # Have the model calculator add all of these blindly to the data |
---|
662 | # tree, and update the graphs which contain them. The fitter |
---|
663 | # needs to be updated to use the I(q) value only, ignoring the rest. |
---|
664 | # |
---|
665 | # The simple fix of returning the existing intermediate results |
---|
666 | # will not work for a couple of reasons: (1) another thread may |
---|
667 | # sneak in to compute its own results before calc_composition_models |
---|
668 | # is called, and (2) calculate_Iq is currently called three times: |
---|
669 | # once with q, once with q values before qmin and once with q values |
---|
670 | # after q max. Both of these should be addressed before |
---|
671 | # replacing this code. |
---|
672 | composition = self._model_info.composition |
---|
673 | if composition and composition[0] == 'product': # only P*S for now |
---|
674 | with calculation_lock: |
---|
675 | _, lazy_results = self._calculate_Iq(qx) |
---|
676 | # for compatibility with sasview 4.x |
---|
677 | results = lazy_results() |
---|
678 | pq_data = results.get("P(Q)") |
---|
679 | sq_data = results.get("S(Q)") |
---|
680 | return pq_data, sq_data |
---|
681 | else: |
---|
682 | return None |
---|
683 | |
---|
684 | def calculate_Iq(self, |
---|
685 | qx, # type: Sequence[float] |
---|
686 | qy=None # type: Optional[Sequence[float]] |
---|
687 | ): |
---|
688 | # type: (...) -> Tuple[np.ndarray, Callable[[], collections.OrderedDict[str, np.ndarray]]] |
---|
689 | """ |
---|
690 | Calculate Iq for one set of q with the current parameters. |
---|
691 | |
---|
692 | If the model is 1D, use *q*. If 2D, use *qx*, *qy*. |
---|
693 | |
---|
694 | This should NOT be used for fitting since it copies the *q* vectors |
---|
695 | to the card for each evaluation. |
---|
696 | |
---|
697 | The returned tuple contains the scattering intensity followed by a |
---|
698 | callable which returns a dictionary of intermediate data from |
---|
699 | ProductKernel. |
---|
700 | """ |
---|
701 | ## uncomment the following when trying to debug the uncoordinated calls |
---|
702 | ## to calculate_Iq |
---|
703 | #if calculation_lock.locked(): |
---|
704 | # logger.info("calculation waiting for another thread to complete") |
---|
705 | # logger.info("\n".join(traceback.format_stack())) |
---|
706 | |
---|
707 | with calculation_lock: |
---|
708 | return self._calculate_Iq(qx, qy) |
---|
709 | |
---|
710 | def _calculate_Iq(self, qx, qy=None): |
---|
711 | if self._model is None: |
---|
712 | # Only need one copy of the compiled kernel regardless of how many |
---|
713 | # times it is used, so store it in the class. Also, to reset the |
---|
714 | # compute engine, need to clear out all existing compiled kernels, |
---|
715 | # which is much easier to do if we store them in the class. |
---|
716 | self.__class__._model = core.build_model(self._model_info) |
---|
717 | if qy is not None: |
---|
718 | q_vectors = [np.asarray(qx), np.asarray(qy)] |
---|
719 | else: |
---|
720 | q_vectors = [np.asarray(qx)] |
---|
721 | calculator = self._model.make_kernel(q_vectors) |
---|
722 | parameters = self._model_info.parameters |
---|
723 | pairs = [self._get_weights(p) for p in parameters.call_parameters] |
---|
724 | #weights.plot_weights(self._model_info, pairs) |
---|
725 | call_details, values, is_magnetic = make_kernel_args(calculator, pairs) |
---|
726 | #call_details.show() |
---|
727 | #print("================ parameters ==================") |
---|
728 | #for p, v in zip(parameters.call_parameters, pairs): print(p.name, v[0]) |
---|
729 | #for k, p in enumerate(self._model_info.parameters.call_parameters): |
---|
730 | # print(k, p.name, *pairs[k]) |
---|
731 | #print("params", self.params) |
---|
732 | #print("values", values) |
---|
733 | #print("is_mag", is_magnetic) |
---|
734 | result = calculator(call_details, values, cutoff=self.cutoff, |
---|
735 | magnetic=is_magnetic) |
---|
736 | lazy_results = getattr(calculator, 'results', |
---|
737 | lambda: collections.OrderedDict()) |
---|
738 | #print("result", result) |
---|
739 | |
---|
740 | calculator.release() |
---|
741 | #self._model.release() |
---|
742 | |
---|
743 | return result, lazy_results |
---|
744 | |
---|
745 | |
---|
746 | def calculate_ER(self, mode=1): |
---|
747 | # type: () -> float |
---|
748 | """ |
---|
749 | Calculate the effective radius for P(q)*S(q) |
---|
750 | |
---|
751 | *mode* is the R_eff type, which defaults to 1 to match the ER |
---|
752 | calculation for sasview models from version 3.x. |
---|
753 | |
---|
754 | :return: the value of the effective radius |
---|
755 | """ |
---|
756 | # ER and VR are only needed for old multiplication models, based on |
---|
757 | # sas.sascalc.fit.MultiplicationModel. Fail for now. If we want to |
---|
758 | # continue supporting them then add some test cases so that the code |
---|
759 | # is exercised. We can access ER/VR using the kernel Fq function by |
---|
760 | # extending _calculate_Iq so that it calls: |
---|
761 | # if er_mode > 0: |
---|
762 | # res = calculator.Fq(call_details, values, cutoff=self.cutoff, |
---|
763 | # magnetic=False, effective_radius_type=mode) |
---|
764 | # R_eff, form_shell_ratio = res[2], res[4] |
---|
765 | # return R_eff, form_shell_ratio |
---|
766 | # Then use the following in calculate_ER: |
---|
767 | # ER, VR = self._calculate_Iq(q=[0.1], er_mode=mode) |
---|
768 | # return ER |
---|
769 | # Similarly, for calculate_VR: |
---|
770 | # ER, VR = self._calculate_Iq(q=[0.1], er_mode=1) |
---|
771 | # return VR |
---|
772 | # Obviously a combined calculate_ER_VR method would be better, but |
---|
773 | # we only need them to support very old models, so ignore the 2x |
---|
774 | # performance hit. |
---|
775 | raise NotImplementedError("ER function is no longer available.") |
---|
776 | |
---|
777 | def calculate_VR(self): |
---|
778 | # type: () -> float |
---|
779 | """ |
---|
780 | Calculate the volf ratio for P(q)*S(q) |
---|
781 | |
---|
782 | :return: the value of the form:shell volume ratio |
---|
783 | """ |
---|
784 | # See comments in calculate_ER. |
---|
785 | raise NotImplementedError("VR function is no longer available.") |
---|
786 | |
---|
787 | def set_dispersion(self, parameter, dispersion): |
---|
788 | # type: (str, weights.Dispersion) -> None |
---|
789 | """ |
---|
790 | Set the dispersion object for a model parameter |
---|
791 | |
---|
792 | :param parameter: name of the parameter [string] |
---|
793 | :param dispersion: dispersion object of type Dispersion |
---|
794 | """ |
---|
795 | if parameter in self.params: |
---|
796 | # TODO: Store the disperser object directly in the model. |
---|
797 | # The current method of relying on the sasview GUI to |
---|
798 | # remember them is kind of funky. |
---|
799 | # Note: can't seem to get disperser parameters from sasview |
---|
800 | # (1) Could create a sasview model that has not yet been |
---|
801 | # converted, assign the disperser to one of its polydisperse |
---|
802 | # parameters, then retrieve the disperser parameters from the |
---|
803 | # sasview model. |
---|
804 | # (2) Could write a disperser parameter retriever in sasview. |
---|
805 | # (3) Could modify sasview to use sasmodels.weights dispersers. |
---|
806 | # For now, rely on the fact that the sasview only ever uses |
---|
807 | # new dispersers in the set_dispersion call and create a new |
---|
808 | # one instead of trying to assign parameters. |
---|
809 | self.dispersion[parameter] = dispersion.get_pars() |
---|
810 | else: |
---|
811 | raise ValueError("%r is not a dispersity or orientation parameter" |
---|
812 | % parameter) |
---|
813 | |
---|
814 | def _dispersion_mesh(self): |
---|
815 | # type: () -> List[Tuple[np.ndarray, np.ndarray]] |
---|
816 | """ |
---|
817 | Create a mesh grid of dispersion parameters and weights. |
---|
818 | |
---|
819 | Returns [p1,p2,...],w where pj is a vector of values for parameter j |
---|
820 | and w is a vector containing the products for weights for each |
---|
821 | parameter set in the vector. |
---|
822 | """ |
---|
823 | pars = [self._get_weights(p) |
---|
824 | for p in self._model_info.parameters.call_parameters |
---|
825 | if p.type == 'volume'] |
---|
826 | return dispersion_mesh(self._model_info, pars) |
---|
827 | |
---|
828 | def _get_weights(self, par): |
---|
829 | # type: (Parameter) -> Tuple[np.ndarray, np.ndarray] |
---|
830 | """ |
---|
831 | Return dispersion weights for parameter |
---|
832 | """ |
---|
833 | if par.name not in self.params: |
---|
834 | if par.name == self.multiplicity_info.control: |
---|
835 | return self.multiplicity, [self.multiplicity], [1.0] |
---|
836 | else: |
---|
837 | # For hidden parameters use default values. This sets |
---|
838 | # scale=1 and background=0 for structure factors |
---|
839 | default = self._model_info.parameters.defaults.get(par.name, np.NaN) |
---|
840 | return default, [default], [1.0] |
---|
841 | elif par.polydisperse: |
---|
842 | value = self.params[par.name] |
---|
843 | dis = self.dispersion[par.name] |
---|
844 | if dis['type'] == 'array': |
---|
845 | dispersity, weight = dis['values'], dis['weights'] |
---|
846 | else: |
---|
847 | dispersity, weight = weights.get_weights( |
---|
848 | dis['type'], dis['npts'], dis['width'], dis['nsigmas'], |
---|
849 | value, par.limits, par.relative_pd) |
---|
850 | return value, dispersity, weight |
---|
851 | else: |
---|
852 | value = self.params[par.name] |
---|
853 | return value, [value], [1.0] |
---|
854 | |
---|
855 | @classmethod |
---|
856 | def runTests(cls): |
---|
857 | """ |
---|
858 | Run any tests built into the model and captures the test output. |
---|
859 | |
---|
860 | Returns success flag and output |
---|
861 | """ |
---|
862 | from .model_test import check_model |
---|
863 | return check_model(cls._model_info) |
---|
864 | |
---|
865 | def test_cylinder(): |
---|
866 | # type: () -> float |
---|
867 | """ |
---|
868 | Test that the cylinder model runs, returning the value at [0.1,0.1]. |
---|
869 | """ |
---|
870 | Cylinder = _make_standard_model('cylinder') |
---|
871 | cylinder = Cylinder() |
---|
872 | return cylinder.evalDistribution([0.1, 0.1]) |
---|
873 | |
---|
874 | def test_structure_factor(): |
---|
875 | # type: () -> float |
---|
876 | """ |
---|
877 | Test that 2-D hardsphere model runs and doesn't produce NaN. |
---|
878 | """ |
---|
879 | Model = _make_standard_model('hardsphere') |
---|
880 | model = Model() |
---|
881 | value2d = model.evalDistribution([0.1, 0.1]) |
---|
882 | value1d = model.evalDistribution(np.array([0.1*np.sqrt(2)])) |
---|
883 | #print("hardsphere", value1d, value2d) |
---|
884 | if np.isnan(value1d) or np.isnan(value2d): |
---|
885 | raise ValueError("hardsphere returns nan") |
---|
886 | |
---|
887 | def test_product(): |
---|
888 | # type: () -> float |
---|
889 | """ |
---|
890 | Test that 2-D hardsphere model runs and doesn't produce NaN. |
---|
891 | """ |
---|
892 | S = _make_standard_model('hayter_msa')() |
---|
893 | P = _make_standard_model('cylinder')() |
---|
894 | model = MultiplicationModel(P, S) |
---|
895 | model.setParam(product.RADIUS_MODE_ID, 1.0) |
---|
896 | value = model.evalDistribution([0.1, 0.1]) |
---|
897 | if np.isnan(value): |
---|
898 | raise ValueError("cylinder*hatyer_msa returns null") |
---|
899 | |
---|
900 | def test_rpa(): |
---|
901 | # type: () -> float |
---|
902 | """ |
---|
903 | Test that the 2-D RPA model runs |
---|
904 | """ |
---|
905 | RPA = _make_standard_model('rpa') |
---|
906 | rpa = RPA(3) |
---|
907 | return rpa.evalDistribution([0.1, 0.1]) |
---|
908 | |
---|
909 | def test_empty_distribution(): |
---|
910 | # type: () -> None |
---|
911 | """ |
---|
912 | Make sure that sasmodels returns NaN when there are no polydispersity points |
---|
913 | """ |
---|
914 | Cylinder = _make_standard_model('cylinder') |
---|
915 | cylinder = Cylinder() |
---|
916 | cylinder.setParam('radius', -1.0) |
---|
917 | cylinder.setParam('background', 0.) |
---|
918 | Iq = cylinder.evalDistribution(np.asarray([0.1])) |
---|
919 | assert Iq[0] == 0., "empty distribution fails" |
---|
920 | |
---|
921 | def test_model_list(): |
---|
922 | # type: () -> None |
---|
923 | """ |
---|
924 | Make sure that all models build as sasview models |
---|
925 | """ |
---|
926 | from .exception import annotate_exception |
---|
927 | for name in core.list_models(): |
---|
928 | try: |
---|
929 | _make_standard_model(name) |
---|
930 | except: |
---|
931 | annotate_exception("when loading "+name) |
---|
932 | raise |
---|
933 | |
---|
934 | def test_old_name(): |
---|
935 | # type: () -> None |
---|
936 | """ |
---|
937 | Load and run cylinder model as sas-models-CylinderModel |
---|
938 | """ |
---|
939 | if not SUPPORT_OLD_STYLE_PLUGINS: |
---|
940 | return |
---|
941 | try: |
---|
942 | # if sasview is not on the path then don't try to test it |
---|
943 | import sas |
---|
944 | except ImportError: |
---|
945 | return |
---|
946 | load_standard_models() |
---|
947 | from sas.models.CylinderModel import CylinderModel |
---|
948 | CylinderModel().evalDistribution([0.1, 0.1]) |
---|
949 | |
---|
950 | def test_structure_factor_background(): |
---|
951 | # type: () -> None |
---|
952 | """ |
---|
953 | Check that sasview model and direct model match, with background=0. |
---|
954 | """ |
---|
955 | from .data import empty_data1D |
---|
956 | from .core import load_model_info, build_model |
---|
957 | from .direct_model import DirectModel |
---|
958 | |
---|
959 | model_name = "hardsphere" |
---|
960 | q = [0.0] |
---|
961 | |
---|
962 | sasview_model = _make_standard_model(model_name)() |
---|
963 | sasview_value = sasview_model.evalDistribution(np.array(q))[0] |
---|
964 | |
---|
965 | data = empty_data1D(q) |
---|
966 | model_info = load_model_info(model_name) |
---|
967 | model = build_model(model_info) |
---|
968 | direct_model = DirectModel(data, model) |
---|
969 | direct_value_zero_background = direct_model(background=0.0) |
---|
970 | |
---|
971 | assert sasview_value == direct_value_zero_background |
---|
972 | |
---|
973 | # Additionally check that direct value background defaults to zero |
---|
974 | direct_value_default = direct_model() |
---|
975 | assert sasview_value == direct_value_default |
---|
976 | |
---|
977 | |
---|
978 | def magnetic_demo(): |
---|
979 | Model = _make_standard_model('sphere') |
---|
980 | model = Model() |
---|
981 | model.setParam('sld_M0', 8) |
---|
982 | q = np.linspace(-0.35, 0.35, 500) |
---|
983 | qx, qy = np.meshgrid(q, q) |
---|
984 | result, _ = model.calculate_Iq(qx.flatten(), qy.flatten()) |
---|
985 | result = result.reshape(qx.shape) |
---|
986 | |
---|
987 | import pylab |
---|
988 | pylab.imshow(np.log(result + 0.001)) |
---|
989 | pylab.show() |
---|
990 | |
---|
991 | if __name__ == "__main__": |
---|
992 | print("cylinder(0.1,0.1)=%g"%test_cylinder()) |
---|
993 | #magnetic_demo() |
---|
994 | #test_product() |
---|
995 | #test_structure_factor() |
---|
996 | #print("rpa:", test_rpa()) |
---|
997 | #test_empty_distribution() |
---|
998 | #test_structure_factor_background() |
---|